Build and validate prediction models for medical decision making. Creates multiple logistic regression models with predicted probabilities that can be directly used in Decision Curve Analysis.
Usage
modelbuilder(
data,
outcome,
outcomePositive,
splitData = TRUE,
randomSeed = 123,
buildBasicModel = TRUE,
basicPredictors,
basicModelName = "basic_model",
buildEnhancedModel = FALSE,
enhancedPredictors,
enhancedModelName = "enhanced_model",
buildBiomarkerModel = FALSE,
biomarkerPredictors,
biomarkerModelName = "biomarker_model",
buildCustomModel = FALSE,
customPredictors,
customModelName = "custom_model",
includeInteractions = FALSE,
interactionTerms = "",
useStepwise = FALSE,
stepwiseDirection = "both",
selectionCriterion = "aic",
transformVariables = FALSE,
transformMethod = "log",
missingDataMethod = "complete_cases",
imputationSets = 5,
crossValidation = FALSE,
cvFolds = 5,
bootstrapValidation = FALSE,
bootstrapReps = 1000,
showModelSummary = TRUE,
showPerformanceMetrics = TRUE,
showCalibrationPlots = TRUE,
showROCCurves = TRUE,
compareModels = TRUE,
createPredictions = TRUE,
exportForDCA = TRUE,
calculateNRI = FALSE,
nriThresholds = "0.05, 0.10, 0.20",
calculateIDI = FALSE,
penalizedRegression = FALSE,
penaltyType = "lasso",
generateRiskScore = FALSE,
riskScorePoints = "simple"
)Arguments
- data
The data as a data frame.
- outcome
Binary outcome variable to predict (e.g., disease presence, adverse event occurrence).
- outcomePositive
Which level of the outcome variable represents the positive case.
- splitData
Split data into training (70\ for unbiased model evaluation.
- randomSeed
Random seed for reproducible data splitting and results.
- buildBasicModel
Build a basic clinical model using demographic and primary risk factors.
- basicPredictors
Variables for the basic clinical model (e.g., age, sex, primary risk factors).
- basicModelName
Name for the basic model (used as column name for predictions).
- buildEnhancedModel
Build an enhanced model with additional clinical variables.
- enhancedPredictors
Variables for the enhanced model (should include basic predictors plus additional ones).
- enhancedModelName
Name for the enhanced model (used as column name for predictions).
- buildBiomarkerModel
Build a model incorporating biomarkers or advanced diagnostics.
- biomarkerPredictors
Variables for the biomarker model (clinical variables plus biomarkers).
- biomarkerModelName
Name for the biomarker model (used as column name for predictions).
- buildCustomModel
Build a custom model with user-specified variables.
- customPredictors
Variables for the custom model.
- customModelName
Name for the custom model (used as column name for predictions).
- includeInteractions
Include two-way interactions between predictor variables.
- interactionTerms
Specify interactions (e.g., "agesex, diabetessmoking"). Leave empty for all pairwise.
- useStepwise
Use stepwise variable selection to optimize models.
- stepwiseDirection
Direction for stepwise variable selection.
- selectionCriterion
Criterion for stepwise variable selection.
- transformVariables
Apply automatic transformations to continuous variables.
- transformMethod
Type of transformation to apply to continuous variables.
- missingDataMethod
Method for handling missing data in predictor variables.
- imputationSets
Number of imputation sets for multiple imputation.
- crossValidation
Perform k-fold cross-validation for robust model assessment.
- cvFolds
Number of folds for cross-validation.
- bootstrapValidation
Perform bootstrap validation to assess optimism and shrinkage.
- bootstrapReps
Number of bootstrap replications for validation.
- showModelSummary
Display regression coefficients and model statistics.
- showPerformanceMetrics
Display AUC, calibration, and other performance measures.
- showCalibrationPlots
Display calibration plots for each model.
- showROCCurves
Display ROC curves for discrimination assessment.
- compareModels
Display side-by-side comparison of all models.
- createPredictions
Add predicted probability columns to dataset for DCA use.
- exportForDCA
Format output specifically for use in Decision Curve Analysis module.
- calculateNRI
Calculate NRI comparing enhanced models to basic model.
- nriThresholds
Risk thresholds for NRI calculation (comma-separated).
- calculateIDI
Calculate IDI comparing models.
- penalizedRegression
Use penalized regression (LASSO/Ridge) for variable selection.
- penaltyType
Type of penalty for regularized regression.
- generateRiskScore
Create integer risk score from best model for clinical use.
- riskScorePoints
Method for creating clinical risk score.
Value
A results object containing:
results$instructions | a html | ||||
results$dataSummary | a html | ||||
results$dcaReadyMessage | a html | ||||
results$basicModelSummary | a table | ||||
results$enhancedModelSummary | a table | ||||
results$biomarkerModelSummary | a table | ||||
results$customModelSummary | a table | ||||
results$modelComparisonTable | a table | ||||
results$validationResults | a table | ||||
results$nriTable | a table | ||||
results$idiTable | a table | ||||
results$riskScoreTable | a table | ||||
results$rocCurvesPlot | an image | ||||
results$calibrationPlotsArray | an array of images | ||||
results$modelComparisonPlot | an image | ||||
results$validationPlot | an image | ||||
results$dcaPreparationSummary | a html |
Tables can be converted to data frames with asDF or as.data.frame. For example:
results$basicModelSummary$asDF
as.data.frame(results$basicModelSummary)